Chiou and Youngs PEER-NGA Empirical Ground Motion Model for ...

Chiou and Youngs PEER-NGA Empirical Ground Motion Model for ... Chiou and Youngs PEER-NGA Empirical Ground Motion Model for ...

peer.berkeley.edu
from peer.berkeley.edu More from this publisher
08.04.2013 Views

the rupture, although this region is mostly to the hanging wall side of a projection of the rupture to the surface (i.e. on the hanging wall of the fault, but the foot wall side of the rupture). It seams reasonable that there should be some increased motion in this area for a buried fault. 1-Rjb/Rcld 1.2 1 0.8 0.6 0.4 0.2 0 -40 -20 0 20 40 60 80 100 120 x (line 1) or y (lines 2 and 3) Line 1 (y=0) Line 2 (x=3) Line 3 (x=10) Figure 17: Variation of the term [ 1− RJB / RRUP ] with location for the three lines shown on Figure 15. Top of rupture is at a depth of 5 km at x=0. The effectiveness of the term [ 1− RJB / RRUP ] to capture the hanging wall effect was tested by fitting the selected PEER-NGA data for peak acceleration with a regression model without a hanging wall term. The residuals were found to be correlated with [ 1− RJB / RRUP ] with a p- n value of 0.0024. Exploratory analysis indicated that the model [ 1− R JB / RRUP ] with n equal to 2 and ½ produced poorer fits to the data (larger standard error) than a linear model (n= 1). No significant difference between normal and reverse faulting data was found (p value of 0.63 for adding a normal fault difference in the correlation). The behavior of the residuals with respect to [ 1− RJB / RRUP ] was checked by separately fitting the data on the hanging wall, 60º ≤ θSITE ≤ 120º, and footwall, -60º ≤ θSITE ≤ -120º sides of the rupture, excluding sites directly above the rupture (RJB=0). Figure 18 shows the residuals and linear fits with respect to [ 1− RJB / RRUP ] constrained to go through 0 at [ 1− RJB / RRUP] =0. The solid lines show the mean trend and the dashed lines the 90% confidence interval for the mean. The residuals on the hanging wall and footwall side of the ruptures have essentially identical trends. Previous empirical models (Abrahamson and Silva, 1997; Campbell and Bozorgnia, 2003) have limited the hanging wall effect to faults with dip angles, δ, less than 70º. Assuming the hanging wall effect is a geometric effect, one might expect that it might correlate with δ. Figure 19 shows residuals for sites on the hanging wall of reverse faults (RJB = 0) plotted versus δ. The data indicate increasing motion with decreasing δ. The trend was modeled with linear functions with 90º - δ, cos(δ) and cos 2 (δ). All three functions show a significant trend with the residuals, with cos 2 (δ) providing a slightly better fit than the other two models. Similar results were found including data from sites with [ 1− RJB / RRUP] > 0.8 and for data from sites with RJB < 5 km. C&Y2006 Page 25

Figure 18: Intra-event residuals for dip-slip earthquake for a model without a hanging wall term. The residuals were fit with a linear regression line constrained to pass through 0 at [ 1− RJB / RRUP ] = 0. Solid line shows the mean trend and the dashed lines indicate the 90% confidence interval for the mean. Figure 19: Intra-event residuals from fitting a model without a hanging wall term for sites with RJB = 0 and FHW = 1 plotted against fault dip, δ. Although, the term [ 1− RJB / RRUP ] appears to model well the hanging wall effect, there are two additional aspects it does not capture. Assuming that the hanging wall effect is a C&Y2006 Page 26

the rupture, although this region is mostly to the hanging wall side of a projection of the<br />

rupture to the surface (i.e. on the hanging wall of the fault, but the foot wall side of the<br />

rupture). It seams reasonable that there should be some increased motion in this area <strong>for</strong> a<br />

buried fault.<br />

1-Rjb/Rcld<br />

1.2<br />

1<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

-40 -20 0 20 40 60 80 100 120<br />

x (line 1) or y (lines 2 <strong>and</strong> 3)<br />

Line 1 (y=0)<br />

Line 2 (x=3)<br />

Line 3 (x=10)<br />

Figure 17: Variation of the term [ 1−<br />

RJB<br />

/ RRUP<br />

] with location <strong>for</strong> the three lines shown on Figure<br />

15. Top of rupture is at a depth of 5 km at x=0.<br />

The effectiveness of the term [ 1−<br />

RJB<br />

/ RRUP<br />

] to capture the hanging wall effect was tested by<br />

fitting the selected <strong>PEER</strong>-<strong>NGA</strong> data <strong>for</strong> peak acceleration with a regression model without a<br />

hanging wall term. The residuals were found to be correlated with [ 1−<br />

RJB<br />

/ RRUP<br />

] with a p-<br />

n<br />

value of 0.0024. Exploratory analysis indicated that the model [ 1−<br />

R JB / RRUP<br />

] with n equal<br />

to 2 <strong>and</strong> ½ produced poorer fits to the data (larger st<strong>and</strong>ard error) than a linear model (n= 1).<br />

No significant difference between normal <strong>and</strong> reverse faulting data was found (p value of<br />

0.63 <strong>for</strong> adding a normal fault difference in the correlation). The behavior of the residuals<br />

with respect to [ 1−<br />

RJB<br />

/ RRUP<br />

] was checked by separately fitting the data on the hanging<br />

wall, 60º ≤ θSITE ≤ 120º, <strong>and</strong> footwall, -60º ≤ θSITE ≤ -120º sides of the rupture, excluding<br />

sites directly above the rupture (RJB=0). Figure 18 shows the residuals <strong>and</strong> linear fits with<br />

respect to [ 1−<br />

RJB<br />

/ RRUP<br />

] constrained to go through 0 at [ 1−<br />

RJB<br />

/ RRUP]<br />

=0. The solid lines<br />

show the mean trend <strong>and</strong> the dashed lines the 90% confidence interval <strong>for</strong> the mean. The<br />

residuals on the hanging wall <strong>and</strong> footwall side of the ruptures have essentially identical<br />

trends.<br />

Previous empirical models (Abrahamson <strong>and</strong> Silva, 1997; Campbell <strong>and</strong> Bozorgnia, 2003)<br />

have limited the hanging wall effect to faults with dip angles, δ, less than 70º. Assuming the<br />

hanging wall effect is a geometric effect, one might expect that it might correlate with δ.<br />

Figure 19 shows residuals <strong>for</strong> sites on the hanging wall of reverse faults (RJB = 0) plotted<br />

versus δ. The data indicate increasing motion with decreasing δ. The trend was modeled<br />

with linear functions with 90º - δ, cos(δ) <strong>and</strong> cos 2 (δ). All three functions show a significant<br />

trend with the residuals, with cos 2 (δ) providing a slightly better fit than the other two models.<br />

Similar results were found including data from sites with [ 1−<br />

RJB<br />

/ RRUP]<br />

> 0.8 <strong>and</strong> <strong>for</strong> data<br />

from sites with RJB < 5 km.<br />

C&Y2006 Page 25

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!